Fun Manager

Nomura Report on Global Memory- 10.19.2026 Nomura's central thesis: The market is overestimating risks surrounding AI financing, slowing AI revenue growth and Chinese memory competition, while underestimating the earnings and cash-generation opportunity of memory manufacturers. They expect memory supply constraints to persist through at least 2028, supporting an extended period of favorable pricing and profitability. → AI infrastructure demand is accelerating. Nomura forecasts: - AI data center capacity grows from 34 GW in 2026 to 74 GW in 2028 — approximately 118% expansion in two years. - AI's share of facility capacity increases from 54% to 76%. - Total global data center power usage rises from 60 GW in 2025 to 115 GW in 2028. - Heavy-agent workloads become an increasingly important source of AI power demand. Nomura considers market concerns surrounding CXMT's progress in 4F² DRAM technology excessive. Its position: - CXMT's technological progress does not invalidate the broader supply-constrained memory thesis. - The breadth of AI demand is being underestimated. - Global memory supply constraints should persist through at least 2028. This is particularly relevant to the scarcity thesis for Micron and SK hynix. They make stock recommendations for Samsung and SK Hynix 🔽 🔽 🔽 Outer Ninth's 3 most important takeaways: 1. The memory scarcity thesis remains intact. Nomura's expectation of tightness through at least 2028 supports the possibility of an extended earnings cycle, rather than a short-lived pricing spike. This is constructive for Micron, SK hynix and broader memory exposure and inline with our views. 2. The separation between infrastructure spending and economic capture is becoming increasingly important. The difference between semiconductor suppliers' cash accumulation and data center operators' indebtedness reinforces one of our core research principles: Participation in AI growth does not necessarily translate into equal financial returns. I've been speaking to you guys more about this recently and I research it thoroughly- debt structure and capital return to shareholders will matter more heading into 2027. Companies that control scarce, essential components may capture a disproportionate share of the economics. 3. For our memory ecosystem, the most important insight is the potential for an extended period of scarcity-driven earnings growth and economic capture by upstream semiconductor suppliers. Companies highlighted on our latest memory ecosystem research w/c I reposted this evening: SK Hynix 하이닉스 000660.KW | $SKHY Samsung 삼성 005930.KW Micron $MU Sandisk $SNDK ESMT 晶豪科 3006.TW Nanya 南亞科 2408.TW Kioxia キオクシア 285A.JP | $KXIAY Winbond 華邦電 2344.TW Phison 群聯 8299.TW AP Memory 愛普 6531.TW Etron 鈺創 5351.TW Seagate $STX Penguin Solutions $PENG Western Digital $WDC Transcend 創見 2451.TW PSMC 力積電 (6770) ETFs- $DRAM & $DISK